Introducing sftrack: A framework for movement data in R

Matthew Boone, Rocio Joo, Mathieu Basille
Ecology Society of America conference 2020

History of movement data


  • Tracking data
    • discrete points
    • home range analysis, resource selection functions, etc…
  • Movement data
    • series of steps
    • Hidden-Markov Models, step selection functions, etc…

History of movement data


  • Proliferation of study in the last 20 years.
  • Devices: gps, satellite, accelerometer, light loggers, cameras, telemetry, etc…
  • Data sizes increasing rapidly with more sensors
figJoo, R., Picardi, S., Boone, M., Clay, T., Patrick, S., Romero-Romero, V., Basille, M., A Decade of Movement Ecology, 2020. arXiv:2006.00110

Software used in the field of Movement Ecology

1Joo, R., Picardi, S., Boone, M., Clay, T., Patrick, S., Romero-Romero, V., Basille, M., A Decade of Movement Ecology, 2020. arXiv:2006.00110

Depends and Suggests of movement packages in R

figJoo, R., Boone, M.E., Clay T.A., Patrick S.C., Clusella-Trullas, S., Basille, M. Navigating through the R packages for movement, Journal of Animal Ecology. 2019. doi: 10.1111/1365-2656.13116

Need for a new defining class

  • Many classes to work with.
  • Methods written in S4.
  • Documentation of classes and methods can be limited.
  • more time = less research

`sf` package

Pebesma, E. Simple Features for R: Standardized Support for Spatial Vector Data, The R Journal. 2019. doi: 10.32614/RJ-2018-009

simple features framework —>

easy mapping and spatial analysis

Goals for `sftrack`

Build a movement framework that the community will buy into

Flexible enough to contain ever expanding structures of movement models

Make methods and documentation transparent

Fully compatible with sf, tidyverse, and ggplot

Goals for `sftrack`

Goals for `sftrack`

Our goals are not to replace any of these existing packages. But to create a class that is flexible enough to be used by all these packages.

Format of an `sftrack`


  • S3, data.frame model
  • sf geometry column
  • grouping class with a required 'id' field
  • time (as.POSIXct or integer)
  • error column (optional)

`sftrack` - reading & inputting

   animal_id                     timestamp       longitude         latitude       fix      
 TTP-053:2753   2018-07-26 15:02:30-04:    4   Min.   :-80.89   Min.   : 0.00   2D  :1978  
 TTP-026:2718   2018-07-30 02:02:30-04:    4   1st Qu.:-80.28   1st Qu.: 0.00   3D  :9570  
 TTP-014:2598   2018-08-01 17:02:30-04:    4   Median :-80.27   Median :26.07   NO  :5517  
 TTP-041:2511   2018-07-25 15:02:30-04:    3   Mean   :-54.89   Mean   :17.83   NULL: 382  
 TTP-002:2264   2018-07-25 18:02:30-04:    3   3rd Qu.:  0.00   3rd Qu.:26.07              
 TTP-058:2258   2018-07-25 20:02:30-04:    3   Max.   :  0.00   Max.   :27.89              
 (Other):2345   (Other)               :17426                                               

devtools::install_github("mablab/sftrack")
raccoon <- read.csv('/home/matt/Downloads/raccoon_data.csv')

summary(raccoon)

Prepare

library(lubridate)
raccoon$timestamp <- ymd_hms(raccoon$timestamp)
# lat/long crs
wsg <- 'EPSG:4326'

`sftrack` - reading & inputting

Create sftrack object

my_sftrack <- as_sftrack(raccoon, burst = 'animal_id', time = 'timestamp', coords = c('longitude', 'latitude'), crs=wsg, zeroNA = TRUE)

head(my_sftrack)

Sftrack with 6 features and 7 fields (1 empty geometries) 
Geometry : "geometry" (XY, crs: EPSG:4326) 
Timestamp : "timestamp" (POSIXct in UTC) 
Burst : "burst" (*id*) 
-------------------------------
      animal_id           timestamp longitude latitude fix         burst                   geometry
3546    TTP-002 2018-04-19 10:02:08 -80.27637 26.07331  3D (id: TTP-002) POINT (-80.27637 26.07331)
3547    TTP-002 2018-04-19 11:02:05 -80.27641 26.07334  3D (id: TTP-002) POINT (-80.27641 26.07334)
3548    TTP-002 2018-04-19 12:02:05 -80.27737 26.07389  3D (id: TTP-002) POINT (-80.27737 26.07389)
3549    TTP-002 2018-04-19 13:02:30 -80.27731 26.07384  3D (id: TTP-002) POINT (-80.27731 26.07384)
15679   TTP-002 2018-04-19 14:02:09 -80.27734 26.07389  2D (id: TTP-002) POINT (-80.27734 26.07389)
11251   TTP-002 2018-04-19 15:02:30   0.00000  0.00000  NO (id: TTP-002)                POINT EMPTY

`sftrack` - plotting works with `sf` methods

plot(my_sftrack, axes = T, graticule = TRUE, pch = 4, lwd=4)

plot of chunk unnamed-chunk-5

`sftrack` - manipulating like a data.frame

my_sftrack <- my_sftrack[c('TTP-058','TTP-041'),]
print(my_sftrack, n_row=12)

Sftrack with 4769 features and 7 fields (2236 empty geometries) 
Geometry : "geometry" (XY, crs: EPSG:4326) 
Timestamp : "timestamp" (POSIXct in UTC) 
Burst : "burst" (*id*) 
-------------------------------
      animal_id           timestamp longitude latitude fix         burst                   geometry
10891   TTP-041 2019-01-15 17:02:30   0.00000  0.00000  NO (id: TTP-041)                POINT EMPTY
3351    TTP-041 2019-01-16 17:02:30 -80.27812 26.06610  3D (id: TTP-041)  POINT (-80.27812 26.0661)
3352    TTP-041 2019-01-17 17:02:30 -80.27830 26.06561  3D (id: TTP-041)  POINT (-80.2783 26.06561)
3353    TTP-041 2019-01-18 17:02:30 -80.27835 26.06566  3D (id: TTP-041) POINT (-80.27835 26.06566)
3354    TTP-041 2019-01-19 00:02:30 -80.27531 26.06607  3D (id: TTP-041) POINT (-80.27531 26.06607)
3355    TTP-041 2019-01-19 01:02:05 -80.27620 26.06573  3D (id: TTP-041)  POINT (-80.2762 26.06573)
3356    TTP-041 2019-01-19 02:02:05 -80.27636 26.06859  3D (id: TTP-041) POINT (-80.27636 26.06859)
10892   TTP-041 2019-01-19 03:02:30   0.00000  0.00000  NO (id: TTP-041)                POINT EMPTY
10893   TTP-041 2019-01-19 04:02:30   0.00000  0.00000  NO (id: TTP-041)                POINT EMPTY
10894   TTP-041 2019-01-19 05:02:30   0.00000  0.00000  NO (id: TTP-041)                POINT EMPTY
15584   TTP-041 2019-01-19 06:02:30 -80.27873 26.06819  2D (id: TTP-041) POINT (-80.27873 26.06819)
10895   TTP-041 2019-01-19 07:02:30   0.00000  0.00000  NO (id: TTP-041)                POINT EMPTY

`sftrack` - manipulating like a data.frame


library(OpenStreetMap)

# Get bounding box from sf geometry
bbox <- st_bbox(my_sftrack)
lower_right <- bbox[c(2,3)] + c(-0.01,0.2)
upper_left <- bbox[c(4,1)] + c(0.01,-0.2)

# download open street map
map <- openmap(upper_left,lower_right,
               zoom=11, type='osm')

# project map
map <- openproj(map)



plot(map)
plot(my_sftrack, add= T, pch=4, lwd=2, col = 'black')

plot of chunk unnamed-chunk-8

`sftrack` - acts just like an 'sf' object


utm_17 <- '+proj=utm +zone=17 +ellps=WGS84 +datum=WGS84 +units=m +no_defs'
my_sftrack <- st_transform(my_sftrack, crs = utm_17)

# Make a polygon within study
poly_pts <- list(
  rbind(
  c(572300, 2883500),
  c(572600, 2883500),
  c(572600, 2883700),
  c(572300, 2883700),
  c(572300, 2883500)
  )
)
polygon <- st_sfc(st_polygon(poly_pts), crs=utm_17)



# What points are within a polygon?

plot(polygon, axes= T, graticule = TRUE, expandBB = c(1.5,1.5,1.5,1.5))
plot(my_sftrack, add =T)

plot of chunk unnamed-chunk-10

`sftrack` - acts just like an 'sf' object


sub_sftrack <- my_sftrack[st_within(my_sftrack, polygon, sparse=FALSE), ]
head(sub_sftrack)

Sftrack with 6 features and 7 fields (0 empty geometries) 
Geometry : "geometry" (XY, crs: +proj=utm +zone=17 +ellps=WGS84 +datum=WGS84 +units=m +no_defs) 
Timestamp : "timestamp" (POSIXct in UTC) 
Burst : "burst" (*id*) 
-------------------------------
      animal_id           timestamp longitude latitude fix         burst                 geometry
3382    TTP-041 2019-01-24 02:02:09 -80.27637 26.07026  3D (id: TTP-041) POINT (572377.1 2883665)
15601   TTP-041 2019-01-28 20:02:30 -80.27576 26.06974  2D (id: TTP-041) POINT (572438.2 2883608)
3425    TTP-041 2019-02-03 19:02:14 -80.27701 26.07041  3D (id: TTP-041) POINT (572313.6 2883681)
3426    TTP-041 2019-02-03 20:02:08 -80.27466 26.07010  3D (id: TTP-041) POINT (572548.8 2883649)
3482    TTP-041 2019-02-13 02:02:09 -80.27630 26.06926  3D (id: TTP-041) POINT (572385.2 2883555)
17035   TTP-041 2019-02-22 02:02:20 -80.27497 26.06999  3D (id: TTP-041) POINT (572517.3 2883636)

`sftrack` - calculate common parameters

step_calc <- step_metrics(my_sftrack)
head(step_calc)

          dx         dy       dist    dt  abs_angle        speed                  sftrack_id
1         NA         NA         NA 86400         NA 0.000000e+00 TTP-041_2019-01-15 17:02:30
2 -18.008768 -53.593639  56.538428 86400 -1.8949668 6.543800e-04 TTP-041_2019-01-16 17:02:30
3  -4.727009   4.625484   6.613601 86400  2.3670495 7.654631e-05 TTP-041_2019-01-17 17:02:30
4 303.425593  47.756375 307.160807 25200  0.1561101 1.218892e-02 TTP-041_2019-01-18 17:02:30
5 -89.014047 -38.261403  96.888779  3575 -2.7356334 2.710176e-02 TTP-041_2019-01-19 00:02:30
6 -17.864152 316.988342 317.491318  3600  1.6270926 8.819203e-02 TTP-041_2019-01-19 01:02:05

summary(step_calc)

       dx                  dy                 dist                dt          abs_angle      
 Min.   :-9821.837   Min.   :-9016.561   Min.   :    0.00   Min.   :  872   Min.   :-3.1390  
 1st Qu.:  -55.167   1st Qu.:  -75.025   1st Qu.:   28.16   1st Qu.: 3584   1st Qu.:-1.5906  
 Median :   -0.312   Median :   -1.547   Median :  125.38   Median : 3600   Median :-0.1664  
 Mean   :  -14.602   Mean   :   -2.238   Mean   :  224.22   Mean   : 5561   Mean   :-0.0418  
 3rd Qu.:   45.378   3rd Qu.:   54.129   3rd Qu.:  288.37   3rd Qu.: 3600   3rd Qu.: 1.5806  
 Max.   : 1700.460   Max.   :14427.086   Max.   :14427.40   Max.   :86400   Max.   : 3.1413  
 NA's   :2793        NA's   :2793        NA's   :2793       NA's   :2       NA's   :2794     
     speed          sftrack_id       
 Min.   :0.00000   Length:4769       
 1st Qu.:0.00000   Class :character  
 Median :0.00000   Mode  :character  
 Mean   :0.03341                     
 3rd Qu.:0.02658                     
 Max.   :7.70282                     
 NA's   :2

`sftrack` - filtering done inside `sftrack` or in other packages


# Filter out by travel distance

my_sftrack <- my_sftrack[!is.na(step_calc$dist) &step_calc$dist<200,]

library(OpenStreetMap)
map_zoom <- 
  openmap(c(26.078,-80.292),c(26.062,-80.262),
          zoom=15, type='osm')
# project to UTM 17
map_zoom <- openproj(map_zoom, utm_17)



plot(map_zoom)
plot(my_sftrack, add= T, pch=4, lwd=2)

plot of chunk unnamed-chunk-15

`sftrack` - sftraj class

my_sftraj <- as_sftraj(my_sftrack)
head(my_sftraj)

Sftraj with 6 features and 7 fields (0 empty geometries) 
Geometry : "geometry" (XY, crs: +proj=utm +zone=17 +ellps=WGS84 +datum=WGS84 +units=m +no_defs) 
Timestamp : "timestamp" (POSIXct in UTC) 
Burst : "burst" (*id*) 
-------------------------------
     animal_id           timestamp longitude latitude fix         burst
3351   TTP-041 2019-01-16 17:02:30 -80.27812 26.06610  3D (id: TTP-041)
3352   TTP-041 2019-01-17 17:02:30 -80.27830 26.06561  3D (id: TTP-041)
3354   TTP-041 2019-01-19 00:02:30 -80.27531 26.06607  3D (id: TTP-041)
3358   TTP-041 2019-01-19 20:02:17 -80.27875 26.07387  3D (id: TTP-041)
3360   TTP-041 2019-01-19 22:02:08 -80.27848 26.07533  3D (id: TTP-041)
3366   TTP-041 2019-01-21 05:02:30 -80.27702 26.06576  3D (id: TTP-041)
                           geometry
3351 LINESTRING (572205.1 288320...
3352 LINESTRING (572187.1 288315...
3354 LINESTRING (572485.8 288320...
3358 LINESTRING (572137.3 288406...
3360 LINESTRING (572163.3 288422...
3366 LINESTRING (572314.8 288316...

plot(my_sftraj, graticule = TRUE, key.pos=4, main = 'Tree Tops Park Raccoons')

plot of chunk unnamed-chunk-17

`sftrack` - dynamic grouping


data('raccoon',package='sftrack')

raccoon$timestamp <- ymd_hms(raccoon$timestamp)

# create a  new month category
raccoon$month <- month(raccoon$timestamp)

burst = c(id = 'animal_id', month = 'month')

my_sftraj <- as_sftraj(raccoon, burst = burst, 
                       time = 'timestamp', coords = c('longitude', 'latitude'),
                       crs=wsg, zeroNA = TRUE)
active_burst(my_sftraj)

[1] "id"    "month"



plot(my_sftraj, graticule = TRUE)

plot of chunk unnamed-chunk-19

`sftrack` - dynamic grouping

active_burst(my_sftraj) <- 'id'
plot(my_sftraj, graticule = TRUE)

plot of chunk unnamed-chunk-20

Core components yet to be installed


  • tibble methods
  • expanded ggplot methods (limited currently)
  • base plot functions need to be expanding to link completely with sf plot methods.
  • Consensus on naming
  • Standards, DBS connections? WKT?

How can you help?


  • Install and try it out! github.com/mablab/sftrack
  • Need input from users + developers
  • Need feedback on final specifications of the framework
    • specifically on sftraj class and grouping structure

Thank You







@birderboone
github/birderboone
mablab.org